At the 6th BEYOND International Technology Innovation Expo (BEYOND Expo 2026) held from May 27 to 30, Zhang Haixing, founder and CEO of humanoid robot company Matrix Super AI, engaged in a discussion. Before establishing Matrix Super AI, Zhang was invited by Elon Musk in 2021 to join Tesla Motors, where he served as the head of the Tesla China Design Center. During the conversation, Zhang emphasized that humanoid robots will eventually enter a phase of comprehensive proliferation, but factors such as strategic choices, cost control, and engineering capabilities will determine which companies succeed in the long run.
The market is sufficiently large to accommodate more than just one Tesla Motors. Prior to founding Matrix Super AI, Zhang Haixing served as the head of the Tesla China Design Center and contributed to the development of the humanoid robot Optimus (later renamed Tesla Bot) and the fully autonomous taxi Cybercab. Zhang acknowledged that Tesla Motors is a highly robust and exceptional company, and his experience there taught him the "first principles" approach. He believes the global robotics market is vast enough to support multiple companies like Tesla Motors, stating, "Chinese companies definitely have opportunities, and more than one will emerge successfully."
When asked whether Matrix Super AI's suppliers overlap with those for Optimus, Zhang noted that many of his company's suppliers are also suppliers for Tesla's electric vehicles. He added that the company benefits from China's mature industrial chain, with a localization rate reaching 85% to 90%, with only some inference chips and high-precision sensor chips relying on imports.
Regarding pricing strategy, Matrix Super AI's Matrix-3 is priced starting at 580,000 yuan, significantly higher than some domestic competitors. Zhang responded that direct price comparisons are not straightforward, as the Matrix-3 is a flagship product with full-size dimensions, 33 degrees of freedom across the body, and equipped with a 27-degree-of-freedom dexterous hand.
Zhang revealed that current orders for Matrix Super AI are scheduled through the end of July 2026, with production in May around 200 units, expected to reach 300 to 400 units in June, and delivery capacity anticipated to achieve 800 to 1,000 units by August, scaling to a mass production volume of 100,000 units in the first half of 2027.
On technological evolution, Zhang introduced Matrix Super AI's classification standards for robot intelligence levels, modeled after autonomous driving: the L1 stage enables remote control operation; L2 involves skills for specific scenarios, such as automatically making tea or selling ice cream; L3 enters the stage of general intelligence, where robots can learn new tasks autonomously with minimal demonstration or verbal instruction, eliminating the need for separate training for each skill. The transition from L2 to L3 is estimated to take about 18 months.
Zhang stated that the L5 stage will represent fully autonomous intelligent partners capable of solving problems without human intervention, a goal expected to take 5 to 8 years to achieve. "Due to the maturity of large AI models, the iteration speed of humanoid robots will be significantly faster than that of autonomous driving in the past, and the progression from L1 to L5 will accelerate. The real bottleneck lies in the supply chain; whoever can reduce costs the fastest and deploy rapidly will gain the advantage."
Like refrigerators and televisions, humanoid robots will become standard household items. Discussing industry trends, Zhang believes the true explosive growth period for the humanoid robot market will occur between 2028 and 2030, with comprehensive proliferation starting after 2030 to 2035. This is primarily constrained by two factors: first, the cost reduction process in the industrial chain, as "costs have not yet dropped to the price level of economical small cars"; second, the generalization capability of the robot's 'brain' remains immature, with "no unified global technical strategy formed yet." Consequently, most robots currently can only adapt to specific scenarios and struggle to switch freely between different tasks.
Regarding application scenarios, Zhang clearly stated that industrial settings are the most practical direction in the short term. However, in the long run, industrial scenarios are not the ultimate destination for humanoid robots. "Industrial clients prioritize cost control, and traditional robotic arms already meet most needs; humanoid robots are actually over-specified for industrial environments."
Zhang is more optimistic about consumer-oriented humanoid robots that possess human-like appearance, dexterous manipulation capabilities, and generalization abilities, enabling them to enter homes and various public service scenarios. He believes this represents the true future of the robotics industry's "ChatGPT moment" or "iPhone moment."
So, when will humanoid robots enter households? Zhang predicts that a batch of lightweight companion chatbots will emerge in the second half of this year to early next year. Within 3 to 5 years, full-size robots equipped with five-fingered dexterous hands and certain generalization capabilities will enter homes, capable of performing household chores like folding clothes and making beds. Ultimately, humanoid robots will become standard household items, much like refrigerators and televisions.
On industry competition, Zhang pointed out that the United States and China each have their advantages: the U.S. possesses top-tier AI talent, leading by about 1 to 2 quarters in the robot's 'brain' aspect, with ample capital support; China holds advantages in engineering, supply chain, and hardware iteration. He believes the industry is not yet in a phase of intense competition, as companies are still at early stages of shipping products and ramping up production capacity.
Zhang forecasts significant industry consolidation within the next 18 to 36 months. "Companies that fail to advance substantive work after fundraising, resting on their capital, and those with unclear strategies engaging in random, trial-and-error R&D, will inevitably face mergers or elimination once capital conditions change and funding becomes unavailable, especially if their own revenue and profits fail to materialize."